[9fans] ape under arm

[9fans] ape under arm

[Jeff Sickel]

Just a quick question for other users of ARM based machines:

Should the ape/lib/openssl be ignored for ARM, or is there a decent alternative for objtype=arm when trying to 
build it?  The current build fails as /sys/include/ape/openssl/opensslconf.h errors out with 'unknown objtype'.

Otherwise, I can get ape/lib/openssl to build (though not necessarily be viable) by the subtle change below:

term% cd /sys/include/ape/openssl
term% diffy opensslconf.h
85c85,86
< #if defined(PLAN9) && defined(T386)
---
> #if defined(PLAN9)
> #if defined(T386)
86a88,89
> #elif defined(Tarm)
> #define ARM_ONLY
89a93,94
> #endif
> 


This small change has made it a lot easier to populate my arm tree.

[9fans] new zlib (1.2.6) support &| build

[Jeff Sickel]

While hitting some ape pcc errors while building my arm tree,
I realized that zlib was quite a bit out of date.  The latest
zlib-1.2.6 can be pulled down and installed in

	/sys/src/cmd/gs/zlib

You'll need to remove gzio.c if you didn't do a complete replace,
and then modify the following:

term% pwd
/sys/src/ape/lib/z
term% diffy *
diff /n/dump/2012/0207/sys/src/ape/lib/z/mkfile ./mkfile
11c11,14
< 	gzio.$O\
---
> 	gzclose.$O\
> 	gzlib.$O\
> 	gzread.$O\
> 	gzwrite.$O\
39c42
< CFLAGS=-c
---
> CFLAGS=-c -D_POSIX_SOURCE -D_C99_SNPRINTF_EXTENSION
diff /n/dump/2012/0207/sys/src/ape/lib/z/z.proto ./z.proto
5a6,9
> arm	- sys sys
> 	lib	- sys sys
> 		ape	- sys sys
> 			libz.a	- sys sys



libz.a is a dependency for quite a few other components, so I'll
post something new after more testing.

-jas

Re: [9fans] new zlib (1.2.6) support &| build

[Jeff Sickel]

don't forget to update /sys/include/ape/zlib.h

Re: [9fans] new zlib (1.2.6) support &| build

[Jeff Sickel]

On Feb 7, 2012, at 3:50 PM, Jeff Sickel wrote:

> don't forget to update /sys/include/ape/zlib.h


And for /sys/src/cmd/gs:

% diffy mkfile src/ld.tr
diff /n/dump/2012/0207/sys/src/cmd/gs/mkfile mkfile
141c141
< 	$CC $CFLAGS zlib/$stem.c; mv $stem.$O obj
---
> 	$CC $CFLAGS -D_C99_SNPRINTF_EXTENSION zlib/$stem.c; mv $stem.$O obj
diff /n/dump/2012/0207/sys/src/cmd/gs/src/ld.tr src/ld.tr
268a269,271
> ./obj/gzlib.o \
> ./obj/gzread.o \
> ./obj/gzwrite.o \


Now gs builds (untested) on/for arm.

Re: [9fans] new zlib (1.2.6) support &| build

[erik quanstrom]

>
> Now gs builds (untested) on/for arm.

gs already built/works on arm.  did someone break it?
it is too slow to use, though.  does anyone know what's
wrong with gs on amd64?

- erik

Re: [9fans] new zlib (1.2.6) support &| build

[Jeff Sickel]

On Feb 7, 2012, at 6:21 PM, erik quanstrom wrote:

>>
>> Now gs builds (untested) on/for arm.
>
> gs already built/works on arm.  did someone break it?
> it is too slow to use, though.  does anyone know what's
> wrong with gs on amd64?

not sure, but it as also failing to build for 386 even
with a recent pull.

-jas

Re: [9fans] new zlib (1.2.6) support &| build

[Jeff Sickel]

On Feb 7, 2012, at 8:48 PM, Jeff Sickel wrote:

>
> On Feb 7, 2012, at 6:21 PM, erik quanstrom wrote:
>
>>>
>>> Now gs builds (untested) on/for arm.
>>
>> gs already built/works on arm.  did someone break it?
>> it is too slow to use, though.  does anyone know what's
>> wrong with gs on amd64?
>
> not sure, but it as also failing to build for 386 even
> with a recent pull.

It's not so much that gs builds/works, but that zlib was
failing to build.  That failure causes ape to fail, and
then a whole slew of other files never get built.

-jas

Re: [9fans] new zlib (1.2.6) support &| build

[Jeff Sickel]

On Feb 7, 2012, at 8:52 PM, Jeff Sickel wrote:

>
> On Feb 7, 2012, at 8:48 PM, Jeff Sickel wrote:
>
>>
>> On Feb 7, 2012, at 6:21 PM, erik quanstrom wrote:
>>
>>>>
>>>> Now gs builds (untested) on/for arm.
>>>
>>> gs already built/works on arm.  did someone break it?
>>> it is too slow to use, though.  does anyone know what's
>>> wrong with gs on amd64?
>>
>> not sure, but it as also failing to build for 386 even
>> with a recent pull.
>
> It's not so much that gs builds/works, but that zlib was
> failing to build.  That failure causes ape to fail, and
> then a whole slew of other files never get built.

aha, technically this was because I had python and some
other contrib packages lumped in w/ my main fossil.

-jas